Dataset¶
- Имаме train set с 25 000 картинки на кучета и котки
- И test set с 12 500 картинки за идентифициране
Нека видим какво имаме в train set-a:
In [172]:
show_cats_and_dogs(0, 6)
Подход¶
- Ще използваме convolutional neural network (конвулационна невронна мрежа?) с keras и tensorflow backend
- С optimizer = RMSprop
- Output layer-a е sigmoid, защото имаме binary crossentropy
Нека видим и base модела:
In [262]:
optimizer = RMSprop(lr=0.0001)
objective = 'binary_crossentropy'
metrics = ['accuracy']
def catdog():
model = Sequential()
model.add(Conv2D(32, (3, 3), padding='same', input_shape=X_train.shape[1:], activation='relu'))
model.add(Conv2D(32, (3, 3), padding='same', activation='relu'))
model.add(MaxPooling2D(data_format="channels_first", pool_size=(2, 2)))
#model.add(Dropout(0.25))
for i in range(6, 9):
model.add(Conv2D(2**i, (3, 3), padding='same', activation='relu'))
model.add(Conv2D(2**i, (3, 3), padding='same', activation='relu'))
model.add(MaxPooling2D(data_format="channels_first", pool_size=(2, 2)))
model.add(Flatten())
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(1))
model.add(Activation('sigmoid'))
model.compile(loss=objective, optimizer=optimizer, metrics=metrics)
return model
In [250]:
model.summary()
Train¶
- 1500 train samples & 500 validation samples
- 20 епохи на CPU
- EarlyStopping за да не overfit-ва
~ 50-60% точност:
In [263]:
model = catdog()
early_stop = EarlyStopping(monitor='val_loss', patience=3, verbose=1, mode='auto')
model.fit(X_train, y_train, epochs=10, validation_split=0.25, shuffle=True, callbacks=[early_stop])
Out[263]:
Да видим какво ще каже моделът и за тестовите данни:
In [265]:
predictions = model.predict(test, verbose=0)
for i in range(0, 10):
if predictions[i, 0] >= 0.5:
print('I am {:.2%} sure this is a Dog'.format(predictions[i][0]))
else:
print('I am {:.2%} sure this is a Cat'.format(1-predictions[i][0]))
plt.imshow(test[i].T)
plt.show()
